Capability Statistics - Cp/Cpk vs Pp/Ppk and Sigma level

[Music] welcome to complexity made simple my name is Paul Allen and the subject of today's newsletter well it's all about capability statistics and what we're going to try and do is to take a look at the full set of capability statistics and how they sort of compare to one another way they're called different names etc one of the things I will say before we start this this is my best attempt to do this this is my understanding I'm going to talk to you also about how I use these capability statistics if you can add anything to this subject and you want to leave some comments in the comments field below then that would be fantastic because it would help us all understand these capability statistics all the more clearly so let's start off we gotta take a look at process capability statistics there's plenty of them about we're going to go CP CP K there's an alternative version called PP PP k so we're going to take a look at those we're going to take a look at those first I guess there's alternative names this one I would call short term capability this one I would call long term capability now CPK PPK essentially what's the difference now the difference is not in the way the data is collected these phrases here this is the short term capability this is the long term capability and what that makes it sound like is it makes it sound like this is how the data was collected so in other words we collected the data over a short period we collected the data over a long period and that's not correct short term and long term is actually about the way the calculation is done and specifically it's about the way standard deviation is calculated so up here the standard deviation this is the shoot methodology and uses the technique r bar over d - what she was he's doing is he's trying to simplify the calculation of standard deviation in the days when he had no calculator and no computer so he's using our bar over d to ever draw a little graph what he's doing he's collecting a subgroup here's the range of the subgroup this is the the maximum value that I saw that's value that I saw and what sureties using is the range so he uses the range for each subgroup so he has a subgroup he has a subgroup he has a subgroup he has a subgroup each one has a range and then he works at the average for that now think about that let's say this is in a so group of 4 samples typically those samples he's put his hand out and he's caught 4 parts of a machine one behind the other so think about that this little subgroup only contains short-term variability whatchu it cannot see is the fact that these subgroups bob up and down so obviously if you worked out standard deviation correctly you would think there was that this much variability if you use chutes methodology you think there is less variability this is known as short term because it is within subgroup and it cuts out lots of the things like changes in material and things like that okay so there is short term long term therefore is the big fat standard deviation calculation every data point take away the average square the difference divided by the number of samples minus one and take the root and that is how PPK and CPK I worked out that is how CP and CPK worked out that's the long term because the long term takes into account everything that's going on okay so they're the basic statistics now there are more statistics to add to this it's add another one see em I see em okay this is known as a it's a machine they talk about it being a machine capability okay what is this thing well this thing is another version of this because what they ask you to do here is to collect 50 consecutive valid 50 consecutive data points in other words what you're trying to do is you're trying to eliminate long term changes due to tool where material changes etc and other adjustments may be due to maintenance deterioration so this is known as machine capability and this is as good as your process capability can literally ever get because it just contains the slight wear and the slight differences in the machine okay so that's CM cmk it's another version of this but just to make it confusing it's using this calculation to calculate the standard deviation but the data has been collected over a short-term period okay and then finally of course we have Sigma level and Sigma capability well Sigma level and Sikma capability Sigma level is essentially ppqe times 3 Sigma capability what's that well it's PP times 3 so it's just another version of this thing now that is my attempt to explain all the process capability measures what I'm going to do next is talk to you about how I use these okay so now we've been through the list of the capability statistics how would you use those which ones would I use which ones do I refer to when do I use them well I'm going to start with the idea of a new product I mean if this probably also applies to when I have a new problem here so it might be a new product I suppose I could say a new problem because the first thing I'm going to do it's the CM cmk in other words what I'm going to do is 50 consecutive config 50 consecutive data points really important hands-off no adjustments let's just make 50 with my hands off the machine and let's see what the machine is capable of and really what I'm asking is do I have a chance in other words is this actual machine is this actual machine capable of doing the job now probably at this point I would probably want a really good capability something with some space here some room because don't forget this variability that hasn't happened yet in this data set therefore I need some room for the process what I'm going to describe it is to get stressed for the long term variables to kick in here okay so I probably want a reasonably good capability now of course if I don't get it the first thing I'm going to do is either change the Machine or look at the machine from a maintenance point of view see if there's something I can do to improve the machine when the bearings need changing maintenance needs doing work that we haven't done for a long long time so I'm going to use CMC m'kay first because if I can't get a decent capability with that when all of this stuff kicks in I really don't stand a chance and then once I've got that then what I'm going to do he's the PP PP k now what I'm going to do is probably collect 50 to a hundred over a number of days maybe hours and I'm going to take a look at what the long-term variability looks like to be honest this is the one that I use the most and then what I do is I think about how I've collected the data so if I collect the data like this over a short period I know that the what variability does it contain you should common sense one favorability does it contain when I open this up and I go for a long term capability I thought it could be days it could be hours I would talk to the people on the machine I talk to the maintenance guys in it so well I've on what period do you expect this to appear and if they say well all the variability would appear in the first three hours then I plate the data over three hours if they say to me well probably take two weeks for all the variables to kick in you know what this might not be weeks or days 30 days or hours this might be weeks so I just used me common sense talk to people get to know your process apply these things to the process don't apply these things based on rules you'll make loads of mistakes when you do that this is a practical thing it will help you to figure out what's going on as long as you're thinking straight and you think okay where did I click the data from it's a long term what does that mean could be that you collect your 50 consecutive you could still be calculating this with a big standard deviation calculation but you know it's got some variability missing you should common sense these tools are fantastic what do they basically do you know they are brilliant because they predict tomorrow why wouldn't you use these things that is what they're for they're to predict the long term defect rates for your process when you've got a new product or indeed a new problem so again I said could be could be product or problem what would I do if I got a problem I want to do 50 consecutive and tzafenat one see how the process works first and I think decide what to do next there are the capability statistics and that is typically the way that I use them the software that I use it's one last thing spc Excel when it does the CPK calculation it's actually doing this one but it's called cbk it's actually doing the long term and that's that's the one that I use there is a version that will calculate the short term as well but when SPC Excel does the CPK for me I'm using this one so if I've collected the data short term I'm doing this if you've collected the date the long term I'm doing this but I keep my wits about me I know what I'm doing if you use these capabilities statistics I'll tell you where the money is disappearing and then you can learn to make more money that's what this stuff is about capability statistics if you'd like to know more about any of the concepts covered in video or any of the other concepts covered in my my other tutorial videos then here's my latest book drink tea and read the paper covers everything you need to know about how to make sure that Six Sigma becomes world class engineering in your company noise if you'd like to get in touch of me a little bit of help about lean a little bit of help with Six Sigma please contact me on the email below [Music]

Channel: Paul Allen

Views: 38,068

Keywords: Pp, Ppk, Cp, Cpk, Capability Statistics, Process capability, https://www.youtube.com/watch?v=KnY8Syl6I3A, Lean, Paul Akers, Shewhart, Control charts, Taguchi, Six Sigma, Lean sigma, DOE, XbarR chart, Gemba academy, Juran, TPS, Khan Academy, Saravanan Kuppusamy, janam sandhu, MIT OpenCourseWare, Professorcleary, Fastcap, Cary Countryman, Doug H, Deming, Shainin, Narendar Sumukadas, Eugene O'Loughlin, statisticsfun, Quality HUB India, Vipin Mandyam, https://www.youtube.com/watch?v=QH984PnwRDE

Id: AGJ1QYI2B4c

Length: 14min 47sec (887 seconds)

Published: Wed May 23 2018